• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

Modulus-86: Composite amplifier achieving <0.0004 % THD+N.

Member
Joined 2009
Paid Member
Links to Modulus-86 build videos on YouTube: Post #2499.

matt_garman went through the Modulus-86 Build Thread and compiled a list of build pictures. You can find his list here: Modulus-86 Build Thread - Post #4463. A nice and compact Modulus-86 build is shown here: Modulus-86 Build Thread - Post #5213.

Getting all the holes in the right spots on the chassis can be a challenge for many. Here's my suggested method: Modulus-86 Build Thread - Post #5215.

Update 10 OCT 2021: Rev. 3.0 released. This revision represents the first major design revision since Rev. 2.0, which was launched in 2015. You can read more about Rev. 3.0 here: Post #2480

Update 16 MAY 2019: Tweaks to the CM cap connections to the THAT1200. The Modulus-86 boards are now made in Canada.

Update 30 DEC 2018: I completely forgot to add Rev. 2.3 to this revision history. That's remarkable as Rev. 2.3 was released nine months ago. :eek: TI finally decided to kill the LME49710, thus Rev. 2.3 uses one of the opamps in the dual LME49720, which is still in current production. The unused opamp is connected as a buffer with its input grounded. Rev. 2.3 uses all current production parts. You can find the product description and measurements here: Neurochrome :: Modulus-86

Update 10 SEP 2017: Modulus-86 is now in Rev. 2.2. TI brought back the DIP version of the LME49710. In Rev. 2.2 I switched back to the DIP version of the IC as some builders found the TO-99 version harder to work with. This reduced the BOM cost by a few bucks too. In addition to this change, the DC servo was updated to reflect the lessons learned from the development of the Modulus-286 while shaving a little more off the BOM cost. The Modulus-86 Rev. 2.2 provides the same stellar performance as Rev. 2.1. To order PCBs, please visit the Modulus-86 Rev. 2.2 product page: Neurochrome :: Modulus-86 Rev. 2.2.

Update 14 APR 2016: Modulus-86 Rev. 2.1 is now available for purchase. The main updates are: Changed LME49710 footprint to TO-99 to address Texas Instruments' decision to discontinue the DIP version. Slight improvement in THD near clipping. Complete rewrite of the chapters on power supply and heat sink dimensioning in the design documentation. See the release announcement in Post #2115.

Update 24 FEB 2015: Modulus-86 Rev. 2.0 is now available offering improved performance. See Post #838.

This spring I set out to design the Mother of All LM3886 Amplifiers. An amplifier so good that it would knock your socks off. But also an amplifier that someone with decent soldering skills could build if provided a well-designed circuit board. This design effort took a while and led to an interesting investigation of the capabilities of the LM3886. It also led me to produce an LM3886 design guide: Taming the LM3886 Chip Amplifier, which I will keep adding material to as it becomes available.

The Modulus-86 is my latest amplifier design. It is a composite amplifier using an LME49710 precision opamp to control an LM3886 power amp. The performance of the resulting composite amp, both in terms of measured performance and in terms of sound quality, is dominated by the performance of the LME49710. Hence, the performance of the Modulus-86 is absolutely stellar. One of the many benefits of the composite topology is that the power supply rejection is very high. This means the Modulus-86 performs as well on a real power supply (I tested with a toroidal transformer, rectifier, and 2x22000 uF) as it does on a regulated, $1200, lab supply (I used an Agilent E3632A). The excellent supply rejection is also one of the main contributors to the stellar sound quality of this amp.
In addition to the LME49710, an OPA2277 precision opamp is used for the DC servo. The use of a DC servo avoids the use of large capacitors in the main signal path.
The input to the amplifier is provided by a THAT1200 differential receiver. This makes it possible to use the XLR connections commonly found in pro audio. The THAT1200 provides performance rivaling that of an input transformer. When using differential connections, this minimizes the induction of hum and EMI, resulting in dead quiet during quiet passages in the music. The differential input can also be configured for use with the single-ended RCA connectors commonly used on prosumer audio gear.

The key features of the Modulus-86 are:
  • 65 W (4 Ω) and 40 W output power (8 Ω) using a Power-86 with the recommended power transformer.
  • Ideally suited for multi-channel amplification.
  • Ultra-low 0.000061 % THD (1 W, 8 Ω, 1 kHz).
  • Ultra-low 0.000067 % THD (40 W, 8 Ω, 1 kHz).
  • Ultra-low 0.00038 % THD+N (40 W, 8 Ω, 1 kHz).
  • Ultra-low 0.000067 % THD (65 W, 4 Ω, 1 kHz).
  • Ultra-low 0.00041 % THD+N (65 W, 4 Ω, 1 kHz).
  • Ultra-low 0.00069 % IMD (40 W, 8 Ω, SMPTE 60 Hz + 7 kHz @ 4:1).
  • 0.065 Hz ~ 85 kHz bandwidth.
  • 14 V/µs slew rate (fully symmetric).
  • 84 kHz full-power bandwidth.
  • 33 µV RMS output noise (20 Hz - 20 kHz, A-weighted).
  • 42 µV RMS output noise (20 Hz - 20 kHz, unweighted).
  • Differential input with 90 dB CMRR eliminating ground loops in the signal path.
  • Phenomenal power supply rejection ensuring consistent, high performance even using unregulated power supplies.
  • Elaborate use of planes and copper pours to maximize circuit performance by minimizing supply and ground impedances.
  • Low-inductance signal ground connects to power ground at one point only for maximum performance.
  • On-board Zobel and Thiele networks for maximum stability even with capacitive loads.
  • On-board EMI/RFI input filter and ESD protection.
  • On-board low noise voltage regulators for the driver op-amp and DC servo.
  • Power and output terminal blocks accept wire sizes up to AWG 10 (5.2 mm2).
  • All leaded. Easy to solder. 90 × 70 mm board footprint.

The THD+N is constant versus frequency throughout the majority of the audio range. The THD+N does rise slightly towards the high end as the loop gain of the amplifier rolls off. At normal listening levels (<1 W), the THD+N is perfectly flat versus frequency, resulting in incredible sound quality.

The differential input makes it super easy to use multiple boards in a bridge, parallel, or bridge/parallel configuration. All that is required to turn a stereo amp into a mono, bridged amp is that two wires are swapped. No component substitutions necessary.

The Modulus-86 has a gain of +20 dB (10x) for better gain structure in the end system. Should a higher gain be desired, the amp can be configured for +26 dB (20x) by changing a resistor. The circuit support the use of the THAT1203 and THAT1206 for gains of +17 dB (7x) and +14 dB (5x), respectively, should lower gains be desired for further optimization of the gain structure.

"But, but ... how does it sound?", I hear you ask. It sounds fantastic! The first time I turned on the Modulus-86 and started the music, I went “WOW!” before even making it to my listening chair. It was immediately obvious that this amp was something special. What struck me was the level of clarity of the reproduction and the deep quiet during quiet passages in the music. Talk about a huge dynamic range! I have played a few instruments in my life – including the trumpet and a brief stint with a drum set. It is especially important to me that metallic instruments (brass wind and cymbal for example) sound metallic and natural. This is an area that challenges many amplifiers and where the Modulus-86 really shines. The midrange is open and natural. The bass is precise and tight. What can I say? I really like it… The detail reproduced from Dire Straits, “Brothers in Arms” and “On Every Street”, for example, is out of this world. I am certain the incredible sonic performance is due to the stellar supply rejection and flat THD+N vs. frequency of the Modulus-86.

For those interested in more background information on composite amplifiers, I suggest reading the article, "Composite Audio Power Amplifiers" in Electronics Now, Nov. 1992 (pp. 38-44). This DIY Audio thread contains a few pictures from the article: composite amplifiers

For the impact of layout on the performance of an LM3886, I suggest looking at this DIY Audio Thread: http://www.diyaudio.com/forums/chip-amps/252436-lm3886-pcb-vs-point-point-data.html

Finally, for a more complete picture, I'll direct you to my Taming the LM3886 page: Neurochrome.com : : Audio : Taming the LM3886 Chip Amplifier

Thanks,

~Tom
 

Attachments

  • MOD86_BlockDiagram.png
    MOD86_BlockDiagram.png
    30.6 KB · Views: 16,902
  • MOD86_R1p0_PCB.jpg
    MOD86_R1p0_PCB.jpg
    404.8 KB · Views: 16,655
  • MOD86_THDvsFREQ_25W_1W_8R.png
    MOD86_THDvsFREQ_25W_1W_8R.png
    168.8 KB · Views: 15,918
  • MOD86_THDvsPOWER_1kHz_8R.png
    MOD86_THDvsPOWER_1kHz_8R.png
    98.1 KB · Views: 15,270
  • MOD86vsLM3886_THD_25W_8R.png
    MOD86vsLM3886_THD_25W_8R.png
    83.4 KB · Views: 15,462
Last edited:
Congratulations - it looks like a continuation of the evolution of nested topologies like the MyRef, MiniRef, etc. We've just started discussion of a DC Servo for the MyRef this year, so it's good to see that you've already done a working implementation here. I hope I can borrow ideas from the DC Servo, if needed.

A few quick questions:

a) Is it possible to omit the differential input and use a single-ended input fed directly to the input of the LME49710 stage without too much surgery?

b) Are alternative opamps like the LME49990 usable without tinkering with the compensation?

c) Is the schematic free for perusal (mainly for the compensation schema)?
 
Also see composite opamp articles by Walt Jung but the one listed above is for lm1875 and obviously the most appropriate reference. Opamp rolling should be easy with no resistor changes.
THAT1200 has SE output so removal should be trivial.
Just my guesses as circuit is not in front of me but composite amps are pretty straightforward and well documented.

Thanks, I've seen some of Walt Jung's composite opamp schematics, and I've also borrowed some ideas from them a few years ago while designing the MiniRef - which uses a Jung-style nested voltage-series feedback loop, but with a Howland current-pump (transconductance block) like the MyRef for the inner high-power chipamp block. It works great, with excellent audible sonics, much like the MyRef, but simpler.

I missed reading the Electronics Now article (referenced above) then, but a quick perusal of the AD711-LM1875 schematic posted at these forums from that article shows that it is significantly sub-optimal - as jcx points out, it throws away a lot of loop gain in a resistive divider in an effort to gain stability (both the MyRef and Miniref are way more sophisticated in applying loop gain for better performance, both locally and globally).

It's possible to implement simple nested voltage/voltage composite opamp/chipamps with fairly trivial (though possibly heavy-handed) dominant-pole compensation - I found one such simulation schematic while digging through some 3-year old simulation files today, which appears to be stable as shown below:

Edit: IIRC, R5 was an attempt to bias the outer opamp into current-sourcing Class-A, which could be omitted or modified with a JFET CCS as needed.
 

Attachments

  • lm1875_composite_acresp.jpg
    lm1875_composite_acresp.jpg
    229.5 KB · Views: 3,970
  • lm1875_composite_expt.jpg
    lm1875_composite_expt.jpg
    89.5 KB · Views: 5,197
Last edited:
Congratulations - it looks like a continuation of the evolution of nested topologies like the MyRef, MiniRef, etc.

The MyRef series is a Howland current pump, which comes with its own unique set of challenges, in particular when it comes to stability. Bob Pease wrote a nice analysis of the circuit in the National Semiconductor Application Note AN-1515 (available as a .pdf from TI): AN-1515 A Comprehensive Study of the Howland Current Pump. It's a pretty easy read, actually.

The MyRef series has several threads dedicated to it. Discussion about the MyRef (including the airing of grievances with its designer) are best served in those threads. I would prefer that this thread remains dedicated to the Modulus-86, though, I welcome general discussion about composite amplifiers and their design challenges here as well.

My goal with the Modulus-86 was to design a circuit, which
  1. Provides the best performance possible.
  2. Uses parts that are current state of the art and commonly available.
  3. Gives the builder the highest possible confidence that he/she will be able to build the circuit and reproduce the performance I measured.

Excellent, Exactly the direction I was heading as well !!! You had put in lots of hard work on it, and, I thank you for it, as well as sharing your knowledge !!!

Thank you very much. It's a route that's well worth taking. This is by far the best semiconductor amp I have designed to date. I've been designing and building amps since the late 1980ies, so that says a bit... :)

a) Is it possible to omit the differential input and use a single-ended input fed directly to the input of the LME49710 stage without too much surgery?

Yep. You can do that. Pull the THAT1200 and connect two pins on the IC footprint with a piece of wire and you have an amp with a single-ended input. You will lose all the benefits of having a differential input, of course. Tradeoffs, tradeoffs...

b) Are alternative opamps like the LME49990 usable without tinkering with the compensation?

The LME49990 is SMD only and I decided early on to develop the circuit using leaded parts for ease of assembly.

The LME49990 has twice the bandwidth of the LME49710, so you will have to tweak the compensation to compensate. In general, I don't recommend opamp "rolling" with composite amplifiers.

c) Is the schematic free for perusal (mainly for the compensation schema)?

In the past, I gave my work away for free. I have decided not to do that this time around.

If you prefer the DIY approach, I suggest starting with the Electronics Now article linked to in Post #1 and my Taming the LM3886 page.

In these types of ultra-high precision circuits, the schematic is only a small part of the circuit. The majority of the circuit performance comes from the PCB itself. Don't believe me? Do what I did: Build an LM3886 amp using a "connect-the-dots" PCB layout. You'll get horrible performance. Optimize the layout and you can beat the data sheet numbers (not by much, but every bit counts). See my LM3886 P2P vs PCB thread for more detail. I'll add some of these observations on my Taming the LM3886 page as time allows as well.

The biggest difference between the Modulus-86 and other composite LM3886 amps is that I operate the LM3886 in closed loop and use the LME49710 for the outer loop controlling the LM3886. I prefer this approach over the more commonly used topology where the LM3886 operates in open loop. My Modulus-86 has much higher phase margin and much better transient response than the more common composite amplifiers where the LM3886 operates in open loop. The clipping behavior is better controlled in the Modulus-86 as well.

Nice ... So, price for 2 90x70mm PCBs is $120 ?

When you buy my boards, you are buying the engineering that went into them. You are buying the confidence that if you assemble the board according to the bill-of-materials, you will be able to match the performance I am measuring. You are buying the piece of mind that comes with knowing that you will receive a well-engineered, well-designed, and fully supported ultra high-end design that delivers state of the art performance. You are also buying the hours of prototyping and testing that went into the board to ensure a successful build once in your hands. The 90x70 mm piece of board material is merely a token representing all the intellectual property of the design.

Also note that the performance of my Modulus-86 is about 100x better than many amplifiers in the >$1k class (comparing THD+N at full power). That's nothing to sneeze at... :)

~Tom
 
Just my guesses as circuit is not in front of me but composite amps are pretty straightforward and well documented.

Well documented, yes. Straight forward, no... :) Once you start scratching the surface, you'll find many pitfalls in composite amps. Stability is the obvious one. That one is pretty straight forward to deal with for someone skilled in the art. Transient behavior, clipping behavior, and graceful recovery from clipping is something quite different. In my previous post, I gave a few clues to how I designed around those pitfalls.

~Tom
 
When you buy my boards, you are buying the engineering that went into them. You are buying the confidence that if you assemble the board according to the bill-of-materials, you will be able to match the performance I am measuring. You are buying the piece of mind that comes with knowing that you will receive a well-engineered, well-designed, and fully supported ultra high-end design that delivers state of the art performance. You are also buying the hours of prototyping and testing that went into the board to ensure a successful build once in your hands. The 90x70 mm piece of board material is merely a token representing all the intellectual property of the design.

Also note that the performance of my Modulus-86 is about 100x better than many amplifiers in the >$1k class (comparing THD+N at full power). That's nothing to sneeze at... :)

~Tom

OK, i go for my second Pass DIY build i think ... he don't charge for development ....

 
Try buying a Pearl board from Papa. Or a B1. The price is okay.

Yeah, Stereo B1 is $30 - one board needed - Pearl is 200 for 2 huge boards with 12 matched jfets - i pay this any day ... Even if you don't like price for pearl,, there is gerber files, and all documentation published online - so you can make what ever you want, how ever you want ...

So, i go for PASS Diy ...

Thing is, this is DIY audio forum, and most of the guys just share projects and support them, like Papa, Salas etc ...

This guy is selling boards, like Elektor etc - not sharing project and this post should be in Vendor forum, not here ...

But this is his right, to ask what ever he want for PCBs , and my right is to say is too pricey in my opinion ...
 
OK, i go for my second Pass DIY build i think ... he don't charge for development ....

"Charge for development"?! You're joking, right? If I was to pay myself the prevailing wage for a design engineer, I'd have to add a couple of zeros to the board price...

I deliver a lot of value and charge accordingly. I deliver more than a piece of fiberglass with holes in it, ya know... I see no problem charging a decent price for a ultra high-end product. A lot of people don't have a problem spending $60 or more per week on fancy Starbucks coffee drinks. The two boards needed for a stereo amp will set these people back two weeks worth of coffee. I think that's pretty inexpensive, actually. Especially considering the performance of my product.

If you want to DIY a circuit similar to mine, I have provided the references for you to do so, and I will be more than happy to help you along the way, for free. However, if you would like me to do the hard work for you, I suggest you purchase one of my boards...

Anyway. I have orders to fill. I should tend to that.

~Tom
 
Tom,

I understand the work that goes into engineering one of these amps. However, it's still a LM3886. On the other hand, I got two PCBs from Rod Elliott for Project 101, for $52 including shipping, a BOM, great support, and documentation galore. This is a beast that pumps out 150W per channel into 8R, almost 300W into 4R, with great sound quality. I'm not knocking your product, just pointing out what's being charged out there in the DIY community. You did a lot of work, and deserve success for that work. From the DIY viewpoint, there's also the question of value as well.

Hong
 
tomchr said:
Ultra-low <0.0004 % THD (25 W, 8 Ω, 1 kHz).

I must be reading LM3886 datasheet wrong ....

Again, i dont care, you can ask what ever you want for your boards, is up to you, but i wont pay that money ... And i am not trying to be rude ...

I pay 180 bucks for Douglas Self preamplifier PCBs, and about the same price for Douglas Self Power Amplifier PCBs, is NOT i DONT WANT TO PAY for someone else work ...

I visit links you posted i read about it, and dunno why you think your "Ultra High Fidelity" amp is better then >$1000 amps ...

I don't prefer theory too much, just a audition, i see you tested with KRK $150 monitors ?!? ... BTW, do you even have this amp built ?

What ever, i am out of this topic ...

 
Tom,

Randomly ran into your thread here. Missed reading the previous work entirely, guess I have to do some reading.

Is the improved distortion spec due to the extra available gain by using the opamp in front of the 3886, increasing the available feedback?

Not sure where the "<0.0004 % THD+N" comes from? Looking at the graphs, the 25w level is substantially higher, in one graph a bit more than that. Also one graph seems to show surprisingly high THD at 1 watt? Am I misreading something?

_-_-
 
I have attached a bunch of measurements of the THD+N for the Modulus-86 and the LM3886 by itself. These measurements were taken using a real supply (toroid transformer + 25 A rectifier + 2x22000 uF reservoir caps). The performance speaks for itself.

The difference between the measured performance of the LM3886 and the data sheet performance is due to the use of a real (unregulated) power supply. The use of a real supply means that the somewhat poor PSRR of the LM3886 becomes a large contributor to the performance. In the composite amp, all this is nulled out by the controlling opamp.

I must be reading LM3886 datasheet wrong ....

No. You're reading the LM3886 data sheet correctly. What you seem to be missing is that this is a composite amplifier where a precision opamp is used to correct for the imperfections of the LM3886. The controlling opamp (LME49710 in my case) is responsible for the stellar distortion numbers and for the superior power supply rejection. The advantages of composite amplifiers are well established in the literature. I've linked to several references in Post #1, I suggest starting with those.

In case of the Modulus-86, you can consider it an LME49710 that is capable of driving a speaker load and deliver up to 11 A of current.

How about some scope shots of overload/clipping behavior and recovery? If you also showed as compared to "normal" LM3886 based designs, even better.

Thank you. I have attached a couple of transient response shots and some comparisons between the LM3886 and the Modulus-86. I'll follow up with the clipping behavior later.

Is the improved distortion spec due to the extra available gain by using the opamp in front of the 3886, increasing the available feedback?

Yep. That's exactly it. The LME49710 control opamp provides an additional 140 dB of loop gain. This is used to drive the imperfections of the LM3886 to, practically, zero.

Not sure where the "<0.0004 % THD+N" comes from?

From the THD+N vs power measurement. The THD+N vs power is done with a 1 kHz test signal and 22 kHz measurement bandwidth. This is high enough bandwidth to include 21 harmonics and all the noise in the audio range. For the THD+N vs frequency, I have to switch to 80 kHz measurement bandwidth to get a realistic measure of the THD at 20 kHz. The additional bandwidth from 22 kHz to 80 kHz adds a bit more noise to the measurement, hence, the THD+N measurement versus frequency will show slightly higher numbers.

Also one graph seems to show surprisingly high THD at 1 watt?

Really? Are you aware of any amp that's better? I'm curious...

~Tom
 

Attachments

  • LM3886_THDvsFREQ_10W_8R.png
    LM3886_THDvsFREQ_10W_8R.png
    87.4 KB · Views: 670
  • LM3886_THDvsFREQ_28W_8R.png
    LM3886_THDvsFREQ_28W_8R.png
    89.1 KB · Views: 880
  • Transient_100nF_8R.jpg
    Transient_100nF_8R.jpg
    250.1 KB · Views: 826
  • Transient_47nF_8R.jpg
    Transient_47nF_8R.jpg
    249.5 KB · Views: 2,947
  • Transient_10nF_8R.jpg
    Transient_10nF_8R.jpg
    245 KB · Views: 3,022
  • MOD86_THDvsFREQ_1W_28V.png
    MOD86_THDvsFREQ_1W_28V.png
    87.3 KB · Views: 3,014
  • MOD86_THDvsFREQ_10W_28V.png
    MOD86_THDvsFREQ_10W_28V.png
    85.5 KB · Views: 3,021
  • MOD86_THDvsFREQ_25W_28V.png
    MOD86_THDvsFREQ_25W_28V.png
    84.8 KB · Views: 3,045
  • LM3886_THDvsFREQ_1W_8R.png
    LM3886_THDvsFREQ_1W_8R.png
    88 KB · Views: 633
Last edited: